Learning From Major Accidents: A Meta-Learning Perspective

The authors investigate meta-learning / transfer learning for accident severity prediction, asking whether knowledge learned from a large, generic accident database can be transferred to a smaller, technology‑specific (and lower-quality) dataset. For PSM incident investigation, this matters because many organizations have limited high-quality local incident data but still need models that generalize. The study suggests a pathway to reuse lessons learned across plants and technologies, reducing the amount of new labeled data needed for useful predictive analytics.